CUDA 11.0 - Hackdot

Re:CUDA 11.0 by ziddoap on Wednesday July 08, 2020 @03:16PM (#23773269)

>cuFFT now accepts __nv_bfloat16 input and output data type for power-of-two sizes with single precision computations within the kernels.

This exact sentence is listed both under "New Feature" and "Known Issues". I'm not super familiar with CUDA stuff, but, it can't be both right?

[ Parent | on HN ]

Re:CUDA 11.0 by flx42_ on Wednesday July 08, 2020 @05:19PM (#23774365)

Thanks, I have reported it internally and it is now fixed.

[ Parent | on HN ]

Re:CUDA 11.0 by usmannk on Wednesday July 08, 2020 @03:19PM (#23773289)

Looks like a mistake, should only be in New Features.

[ Parent | on HN ]

Re:CUDA 11.0 by bjornsing on Wednesday July 08, 2020 @04:10PM (#23773740)

So a known issue in the known issues?

[ Parent | on HN ]

Re:CUDA 11.0 by willwill100 on Wednesday July 08, 2020 @06:31PM (#23774932)

A known known

[ Parent | on HN ]